Sunteți pe pagina 1din 5

Cloud Agency- A Mobile Agent Based

Cloud System
1
Srinivasan.R,
2
Srihariprasath.B,

!airamoorthy.",
#
Bas$aran.%
1
Asst professor,
2, 3,4
PG students
Department of Information technology,
Velammal Engineering ollege, hennai, India
sanjaisrini@gmail.com, srihari.velammal@gmail.com, successvairam@gmail.com
AbstractThe cloud paradigm appeared on the
computing scene in 2005 with the Amazon Elastic
Compute Cloud (EC2). After this date a large set
of related technologies has !een de"eloped. #n the
academic world and especiall$ in the %&C area
cloud computing is in some wa$ in competition
with the '(#) model which offers a middleware
!ased approach. *ne of the solutions proposed is
the integration of the two paradigms in order to
use the enormous potential of the e+istent
computational '(#)s in new wa$s. *ne of the
most diffused pro!lems on these s$stems is the
choice of the correct programming paradigm,
man$ different approaches e+ist and it is difficult
to define which is the approach that !est fit with
the cloud paradigm. #n this paper we propose the
integration of a Cloud on '(#) architecture with
a mo!ile agent platform. The architecture we
propose offer -irtual clusters with full
administrati"e control to final users adopting an
e+istent '(#) architecture and especiall$ its
securit$ infrastructure. The mo!ile agent
platform is a!le to d$namicall$ add and configure
ser"ices on the "irtual clusters. The e+perience
here presented shows that the mo!ile agent
paradigm well fulfills the d$namic properties of
the Cloud paradigm and could !e a good choice to
simpl$ de"elop application and ser"ices a!le to
d$namicall$ adapt themsel"es to the "irtualized
en"ironment.
.e$wordsCloud Computing '(#) /o!ile
Agents.
&. &"'R()*C'&("
'he +merging Cloud Computing
paradigm, due to an intensive use o, the
virtuali-ation approach, o,,ers to users resources on
.hich they have ,ull administrative control. Cloud
systems o,,er services at t.o di,,erent layers/
&n,rastructure 0ayer that provides management
services able to migrate 1or move, modi,y2 virtual
machines3 !irtual Machine 0ayer that provides
di,,erently pre-con,igured !irtual Machines. *sually
the &n,rastructure 0ayer is designed according
to the Service (riented Architecture 1S(A2 model,
e4ploiting .eb services technology. 'he !irtual
Machine 0ayer shares a set o, virtual machine
images, .hose management
depends on the adopted virtuali-ation engine. A,ter
the Ama-on +C2 plat,orm 516, in ,e. years, a great
number o, pro7ects have started to build up ne.
Cloud systems, both on commercial environments
1Microso,t A-ure 526 and &BM Cloud Computing
562, and open scienti,ic plat,orms 1"imbus and the
virtual 8or$spaces5#-962 ore research pro7ects
5Reservoir 5:62. At the state o, the art Cloud
Computing systems are logically subdivided into
three di,,erent $inds/
&aaS &n,rastructure as a Service, i.e. systems .hich
o,,er a ,ully virtuali-ed architecture based on a
service oriented inter,ace, as an e4ample it is
possible to create and deliver a 1virtual2 machine to a
,inal user leaving him ,ully administrative control.
;aaS ;lat,orm as a Service, i.e. the system o,,ers a
virtuali-ed plat,orm, usually .ith a .eb inter,ace, on
.hich it is possible to develop applications,
customi-e systems,
AaaS Application as a Service, i.e. the system hosts
applications that can be accessed directly through a
.eb inter,ace 1as an e4ample <oogleApps2.
&n academic .orld Cloud computing is going to be
integrated .ith the <R&) conte4t, .hich collects
large computational
Resources in a single in,rastructure. Solutions li$e
526, 56, 5#6 aim at trans,orming the actual =;C
resources
in Cloud &n,rastructures dedicated to deliver
virtuali-ed in,rastructures 1&n,rastructure as a
Service2.
&n this paper .e propose architecture3 .e named
Cloud Agency .hich aims at o,,ering !irtual
Clusters on the top o, an e4isting <R&) that can be
easily con,igured .ith the support o, mobile agents
based services.
Mobile Agents 51>6 are the last evolution o,
Mobile Code based systems. 'hey add mobility to a
set o, ,eatures, such as reactivity, proactivity,
communication and social ability, .hich characteri-e
ordinary so,t.are agents systems.
A mobile agent is a program, .hich is able
to migrate across the net.or$ bringing its o.n code
Proceedings of !ational onference on Ad"anced omputing and ommunication#!A11, April$1%, 2&11
and e4ecution state. 'he idea o, a sel,-controlled
e4ecution that is able to migrate close to data source
.as proposed by 511,126 as the ne. .ave to replace
the client-server paradigm .ith a better, more
e,,icient and ,le4ible mode o, communication. Many
.or$s, .hich deal .ith design o, agent based <rid
plat,orms 5126, use mobile agents technology in order
to implement basic services addressing some topics
such as load balancing, ,ault mas$ing and service
discovery 516. &n 51#6 so,t.are agents are utili-ed in
order to manage and assign resources and to
distribute applications and data.
A mobile agents based approach .ill
allo. to move services through di,,erent virtual
machines and in di,,erent cloud conte4ts. A set o,
speciali-ed agents .ill enrich the plat,orm .ith a set
o, ,acilities ,or per,orming measures about the user-
perceived per,ormances and to dynamically adapt the
allocation o, resources on the basis o, the e,,ective
load and the actual service level o,,ered by a cloud.
'he remainder o, this paper is structured as
,ollo.s. "e4t Section & o,,ers an overvie. o, the
proposed architecture, Section && describes the
clouding in,rastructure .e built3 section &&& describes
the mobile agent plat,orm and its inter,ace.8ith
<R&) services, &! illustrates a typical architecture
use case.
&&. '=+ C0(*) A<+"C? ARC=&'+C'*R+
'he architecture .e propose aims at
integrating three di,,erent technologies/ <R&), Cloud
and Mobile Agents. 'he speci,ic role o, <R&) in the
proposed architecture is to o,,er a common, .ell-
$no.n plat,orm, =igh ;er,ormance Computational
resource, together .ith a stable and .ell accepted
security in,rastructure. 'he Cloud ;aradigm
,urnishes standards and technologies in order to
distribute virtual Clusters to ,inal users, leaving them
,ull administrative control over an easily scalable
resource. 'he Mobile Agents capabilities introduce
,eatures that provide the ,inal users a simple and
really ,le4ible .ay to con,igure the available !irtual
Clusters.
@igure 1. 'he Cloud Agency Architecture
@igure.1 outlines the approach .e propose.
'he architecture ,ounds on a !irtual Cluster Cloud
architecture, .hich aims at o,,ering !irtual clusters
on the top o, a <R&) in,rastructure, this architecture
is described in detail in section &&.
Synthetically it o,,ers a service 1!CService2
.hich accepts as an input an AM0 cluster
description, builds the !C on the available physical
resources, and returns a re,erence to the cluster. 'he
virtual Cluster obtained is con,igured in order to host
a <lobus container 1<C2 .ith a valid certi,icate, the
mobile agent plat,orm and its <R&) services
inter,ace 1described in detail in section &&&2. 8hen the
!C starts up it, due to the certi,ied <C, is available
as a resource
in the <R&) in,rastructure and the precon,igured
MA<)A Services can be invo$ed directly ,rom the
users. "ote that
the mobile agent plat,orm is accessible only through
the <S inter,ace, and so its access is granted using
the underlying <R&) security system.
'he Cloud Agency plat,orm o,,ers an
additional service, Repository, .hich can be invo$ed
by the client in order to recover agent-based services
and, through the MA<)A services, to migrate the
service on the target cluster.
'he client hosts a Cluster Con,igurator,
.hich is able to Buery the Repository service in order
to obtain the list o, the available agent based services
and to launch them on the target cluster. 'he !C
con,iguration can ta$e place both at !C startup and
on-the-,ly, i.e. .ith the !C already running, so
o,,ering services .ithout stopping the system.
&&&. '=+ !&R'*A0 C0*S'+R BAS+)
C0(*) ARC=&'+C'*R+
As mentioned be,ore, our solution builds
a &aaS 1&n,rastructure as a Service2 cloud
environment upon a <R&) in,rastructure. (ur model
o, the in,rastructure is a collection o, clusters, each
o, .hich is composed o, a ,ront-end node 1@+2 and a
set o, computing nodes e4changing messages on a
private net.or$. Both the nodes and the net.or$ can
be physical or virtual. &n its current implementation,
.e do not support hybrid clusters, i.e., clusters made
up o, a mi4ture o, physical and virtual nodes.
&n the ,ollo.ing .e .ill use the term cluster to re,er
to a generic cluster, speci,ying i, it is virtual or
physical only .hen necessary.
'he clusters participate in the underlying <R&) and
o,,er their computational resources to the <R&)
in,rastructure. 'heir @+s host a <lobus container and
are certi,ied .ithin the <R&) !irtual organi-ation.
'he @+s also hosts the MA<)A mobile agent
plat,orm and the MA<)A <R&) services 1&&&2.
@igure 2 describes the overall architecture. 'he
application client resides on a user machine 1.hich
has access to the <R&) environment2 and interacts
.ith the system through invocation o, <R&)
services. @urthermore, it manages <R&) connections,
15
Adhiparasakthi Engineering College, Melmaruvathur
also providing utilities ,or end-users 1notably,
per,ormance analysis services2. 'he architecture
provides <R&) services that enable the user to build
up a ne. cluster
as a <R&) !irtual 8or$space 5C6, 5D6 .ith ,ull
access rights.
@igure 2. 'he !irtual Cluster architecture

'he !irtual Cluster architecture &n light o, the
above, the architecture can be subdivided into three
main components, as is sho.n in above ,igure.
Services, .hich o,,er the ,unctionalities to
the <R&) environment. 'he component
implementation relies on a <R&) service,
named !CService
&mages, .hich are the !irtual Cluster "ode
images, containing all the so,t.are needed
to integrate the !C into the <R&)
environment 1a <' container2 and the
mobile agent plat,orm 1MA<)A2
Client, .hich allo.s the ,inal user to
interact .ith the Cloud environment.
'he proposed architecture is shared .ith another
e-science cloud pro7ect .e are carrying on,
;er,Cloud 5E6, .hich is a
@rame.or$ that provides per,ormance prediction
services in an e-science cloud.
&!. M(B&0+ A<+"' 8&'= <R&) ACC+SS
&n 596 .e de,ined a model o, agent-enabled <rid
node, in the ,ollo.ing .e outline the architecture .e
have designed and reali-ed. &t allo.s to develop and
deploy agent based
Services in a <rid middle.are. An agent-enabled
<rid node is composed o, a Mobile Agents plat,orm
distributed on di,,erent machines .hich are
connected by an internal
"et.or$ such as a cluster, and o, a so,t.are
component that ma$es available mobile agents
,acilities to authenticated and
authori-ed <rid users only by some de,ined <rid
services.
'he above ,igure 1a2 sho.s ho. a client
can use a 8SR@ stub to invo$e agent based <rid
services. A so,t.are component, to be installed on a
<rid ,ront-end, acts as a pro4y bet.een the <rid
technology and a Mobile Agent plat,orm, .hich can
be distributed internally. An agent plat,orm is
composed o, agent containers. A container is a peer-
to-peer application that supports agentsF e4ecution.
Containers can e4ecute on the same machine or on
di,,erent machines. 8e .ill call agent-node, a
machine that hosts at least an agent cont ainer.
&n the Cloud Agency Architecture all the !C
nodes can be agent-nodes. Any $inds o, agent service
can be built on top o, this in,rastructure and can be
used by a <rid user, .ho is una.are to e4ploit agent
technology. &n 56 .e presented a ,irst set o, agent
based <rid services to support monitoring and
management o, mobile agent plat,orms and
applications e4ecuting on a <rid resource.
8e have built MA<)A as a <lobus <'#
plug-in by e4tending and integrating so,t.are open
sources. 8e used Gade technology to enable agent
e4ecution on our in,rastructure. Gade represents a
@&;A compliant implementation o, a Mobile Agent
plat,orm. &t provides the environment
to host agents e4ecution and some basic ,acilities,
.hich characteri-e the agent li,e cycle, and those
services, .hich are mandatory in a @&;A compliant
architecture. &n particular )@ 1)irectory @acilitator2
is an agent that implements a registry to publish and
discover services provided by other
agents.
1a2
1b2
@igure . Agent enabled <rid node
AMS is an agent provides administration
services to manage plat,orm and agents. @inally, in
order to manage the plat,orm .ith a service inter,ace
.e adopted 8SA<, a ,rame.or$ that implements the
8SA& 18eb Services Agent &ntegration2 pro7ect. &t
supports the communication bet.een 8eb Services
and agents. 8SA< allo.s communicating .ith a
16
Proceedings of !ational onference on Ad"anced omputing and ommunication#!A11, April$1%, 2&11
Gade agent by a 8eb Services inter,ace. 8e adapted
the ,rame.or$ in order to support 8SR@ technology.
'he distribution o, architectural components o, an
Agent.
<rid node on the available machines is
sho.n in @igure b. (n the 1virtual2 cluster ,ront-end
is installed both the <lobus container and a Gade
Main Container. 'he <lobus container e4ports 8eb
Services by .hich it is possible to invo$e the
deployed agent services or to start ne. containers
and agents on a .or$ing machine.
'he Gade Main Container manages
containers running on the other nodes belonging to
the <rid resource. &t hosts )@ and AMS. Containers,
.hich are running on other machines, ta$e part to the
agent plat,orm o, the Agent <rid "ode. +ach
computing node is connected .ith the others via a
local private net.or$. (nly the ,ront-end has direct
access to the e4ternal net.or$.
@our <rid Services have been deployed into
the <lobus container. 'hey provide an inter,ace by
.hich it is possible to access the agent plat,orm. @or
Monitoring, Management, and Application 8SA<
technology allo.s to as$ AMS and )@ ,or retrieval
o, available in,ormation about the current status o,
agent plat,orm and to manage agents and containers.
&n the same .ay it is possible to communicate .ith
other agents running on the plat,orm in order to
e4ploit their services.
!. !&R'*A0 C0*S'+R )?"AM&C
C("@&<*RA'&("
'he !C based cloud architecture outlined in
section &&, o,,ers customi-ed clusters to users and
maintains a large repository o, cluster virtual images.
(nce a virtual cluster is delivered to the ,inal user,
according to the speci,ied con,iguration, i.e. a,ter
installation o, additional so,t.are and ,ine tuning o,
operating system, it can be released to under the ,ull
the userFs control. @rom user perspective there is no
di,,erence bet.een a physical and a virtual cluster.
+ven i, the advantages o, the approaches are
clear, ne. problem arise/ ho. to simpli,y the !C
customi-ation ,or the usersH =o. to add ne.
,unctionalities to the !C in a simple and clear .ay
.ithout modi,ying the behavior o, already .or$ing
servicesH =o. it is possible to scale a con,igurationH
Adoption o, the Mobile Agent plat,orm is the
solution .e propose to the problem/ it let a user to
choose bet.een a set o, available services, and
migrate them on the target !C.
'he Cloud Agency )ynamic Con,iguration
module is composed o, a set o, agent-based services
and a Repository 1<R&)2 service, .hose methods
returns the list o, available services and let the user
obtain them in order to upload the services on the
target !C. 'he MA<)A-<R&) integration that has
been described above lets us to access to the plat,orm
services trough the <R&) in,rastructure and its
security services, granting the security o, the
proposed solutions/ only a certi,ied user is able to
move agents into the !C. &n order to better describe
the con,iguration description .e propose a simple
usage e4ample, in .hich .e add a 8eb server Agent
into a !C, this is a toy e4ample, .hich only aims at
sho.ing ho. the architecture .or$s.
@igure # brie,ly summari-es the complete
procedure. 'he Cloud Agency Client 112 send a <S
reBuest to the Repository <S obtaining 122 the list o,
the available services. A,ter that the user chooses the
one he is interested on and send a ne. reBuest 12 to
the <S, in order to retrieve it 1#2and then start it on
the !irtual Cluster 1C2. 'he MA<)A <S starts up
the agent into the plat,orm 1D2 and ends up the
procedure.
@igure #. 'he Cloud Agency Architecture
At this point the !C is con,igured and the
ne. service 1in the e4ample a .eb server2 is
available on the !irtual Cluster. 'he MA<)A-<R&)
integration o,,er additional services that enable the
,inal user to clone or move on one o, !C node the
agent-based services. &t is important, then to point out
that the Repository Services, in the above e4ample, is
deployed into the <lobus containers hosted on the
;hysical Cluster .hich o,,ers the !CService, but it
can be hosted in every container into the <R&)
in,rastructure, hosted by virtual or physical
machines.
!&. R+0A'+) 8(R% - C0(*)
'+C="(0(<&+S
'he cloud paradigm appeared on the
computing scene in 2>>C .ith the Ama-on +lastic
Compute Cloud 1+C22 516. A,ter this date, a large set
o, related technologies has been developed. &n
commercial conte4ts, it is .orth mentioning the
&BMs Blue Cloud 5:6, the Sun Microsystems
"et.or$.com 51>6, the Microso,t A-ure Services
;lat,orm 5116, the <oogle App +ngine 5126, the )ell
Cloud computing solutions 516. Most o, these
commercial systems adopt proprietary solutions
1such as the virtuali-ation engine by !M8are 51#62,
and relatively ,e. details are available on the
adopted architectures. &n the academic .orld, and
especially in the =;C area, cloud computing is in
17
Adhiparasakthi Engineering College, Melmaruvathur
some .ay in competition .ith the <R&) model, as
outlined by Gha, Mer-$y and @o4 51C6.
'he idea o, <R&)-Cloud integration and the
adoption o, virtuali-ation techniBues in <R&)
in,rastructure .as e4plored in research pro7ects as
Reservoir 5#6, and in technologies as open"ebula 56
and virtual .or$spaces 5D6, 51D6, .ith the derived
cloud tool$it "imbus526. At the state o, the art,
e4amples o, e-science clouds are Beginning to
emerge 51E6, 526, 5196, 51:6. 'hey are based (n the
above-mentioned technologies and have architectures
Similar to the one presented in this paper, even i,, at
the Best o, the authorsF $no.ledge, none o, them
provides ;er,ormance evaluation and prediction tools
as services &ntegrated in the architecture.
A number o, activities about applying agents in
computational <rids have been initiated in past years
52>6. An e4ample (, an agent <rid is presented in
5216 and 5226. A <rid based Architecture ,or
Multimedia services management is ;resented in
526. 52#6 )e,ines an architecture o, a services
)iscovery ,unctionality implemented through mobile
agent 'echnology. &n 52C6 so,t.are agents are
employed in order to manage and assign resources
and to distribute applications And data. A similar
architecture is described in 52D6 .here all the ,eatures
o, 7ob migration .ill be encapsulated in a user
program .rapper that is implemented on a Gava layer
Bet.een a mobile agent and the corresponding user
program. 52E6 and 5296 also ,ocus on monitoring and
management o, <rid resources. @inally some other
challenges dealing .ith <R&) and mobile agents
programmings are addressed such as security 52:6
and e4ploitation o, <R&) capability ,rom
=eterogeneous and mobile devices 5>6. 8e
proposed in 516 a <rid in,rastructure ,or agents
based <rid computing, but it .as not integrated in a
<rid middle.are yet. @urthermore it did not o,,er
<rid services by standard inter,aces. &n 526 <lobus
services have been used to enhance agents
capabilities And to allo. their e4ecution on <rid
in,rastructure. 'hese mechanisms allo.ed
transparent migration o, agents among <rid nodes,
but did not provide the possibility to develop ne.
<rid services by agent technology.
!&&. C("C0*S&("
&n this paper .e have presented the
architecture o, Cloud Agency, .hich o,,ers cloud-
on-<R&) ,unctionalities integrated
.ith a mobile agent plat,orm able to dynamically add
and con,igure services on the virtual clusters. 'he
architecture o, Cloud Agency ma$es use o, e4isting
<R&) and virtuali-ation technologies to manage at
lo.level the virtual clusters. &n order to build up the
dynamic con,iguration system .e adopt a mobile
agent plat,orm enriched, .ith a <R&) service
inter,ace, and integrates it in the available !irtual
clusters. Moreover .e also provide a dedicated set o,
services able to collect agent-based services.
'he main contribution o, our .or$ is undoubtedly
the possibility to dynamically con,igure the !C
received ,rom the cloud, in a simple manner
transparent to the user. 'his is o, great importance
,or end-users, because it ma$es it possible to manage
the o,,ered services, .ithout changing the underlying
!irtual Cluster con,iguration. 'he e4perience than
demonstrated us that the mobile agent programming
paradigm .ell suit to the Cloud paradigm, because
the agent-based applicationIservices easily adapt
themselves to dynamically changing systems. As
regards the ,uture evolution o, our .or$, .e aims at
per,orming a clear evaluation o, the per,ormances o,
the proposed solutions, in order to have a clear
evaluation o, the usability o, the architecture. 'he
,ollo.ing really relevant step is the improvement o,
the Repository service, in order to build up a ,le4ible
system ,or search and management o, agent-based
<R&) services. Moreover .e are .or$ing in
integrating the architecture .ith an engine able to
dynamically discover services on the basis o, their
semantic, using a graph matching approach 1see
562. &n parallel .ith the above described tas$s, .e
aims at e4tending as much as possible the set o,
agent-based <S to be o,,ered to ,inal users.
!&&&.R+@+R+"C+S
516 Ama-on &nc., J+lastic compute cloud,K "ov 2>>9,
http/IIa.s.ama-on.comIec2.
526 *niversity o, Chicago, J"imbus pro7ect,K 2>>:,
http/II.or$space.globus.orgIcloudsInimbus.html.
56 )istributed Systems Architecture Research <roup,
J(pennebula pro7ect,K *niversidad Complutense de Madrid, 'ech.
Rep., 2>>:, http/II....opennebula.org.
5#6 R. Consortium, JReservoir pro7ect,K 2>>:,
http/II...>.ibm.comIpressIusIenIpressreleaseI2##9..ss.
5C6 &. '. @oster, '. @reeman, %. %eahey, ). Sche,tner, B.
Sotomayor, and A. Lhang, J!irtual clusters ,or grid communities,K
in CC<R&). &+++ Computer Society, 2>>D, pp.C1MC2>.
5D6 %. %eahey, &. '. @oster, '. @reeman, and A. Lhang, J!irtual
.or$spaces/ Achieving Buality o, service and Buality o, li,e
in the grid,K Scienti,ic ;rogramming,
5E6 +. ;. Mancini, M. Ra$, and *. !illano, J;er,cloud/ <rid
services ,or per,ormance-oriented development o, cloud
computing applications,K in ;roc. o, +merging 'echnologies ,or
"e4t generation <R&) 1+'"<R&)-2>>:I8+'&C+-2>>:2,.
596 R. A!+RSA, B. ). MAR'&"(, and S. !+"'&C&"N*+,
J&ntegration o, mobile agents technology and globus ,or
assisted design and automated development o, grid services,K
in 12th &+++ &nternational Con,erence on Computational
Science and +ngineering 1CS+ 2>>:2,.
5:6 &BM &nc., JBlue cloud pro7ect,K Gune 2>>9,
http/II...>.ibm.comIpressIusIenIpressreleaseI22D1..ss.
51>6 Sun Microsystems, J"et.or$.com,K
http/II....net.or$.com.
5116 Microso,t Co., JA-ure services plat,orm,K
http/II....microso,t.comIa-ureIde,ault.msp4.
5126 <oogle &nc., J<oogle application engine,K
http/IIcode.google.comIintlIit-&'Iappengine.
18

S-ar putea să vă placă și